Merge pull request #131 from spotify/bgenchel/add-medleydb

added medleydb dataset, test file, and include in download

basic_pitch/data/datasets/guitarset.py

@@ -136,17 +136,19 @@ def create_input_data(
     if seed:
         random.seed(seed)
 
-    def determine_split() -> str:
-        partition = random.uniform(0, 1)
-        if partition < validation_bound:
+    def determine_split(index: int) -> str:
+        if index < len(track_ids) * validation_bound:
             return "train"
-        if partition < test_bound:
+        elif index < len(track_ids) * test_bound:
             return "validation"
-        return "test"
+        else:
+            return "test"
 
     guitarset = mirdata.initialize("guitarset")
+    track_ids = guitarset.track_ids
+    random.shuffle(track_ids)
 
-    return [(track_id, determine_split()) for track_id in guitarset.track_ids]
+    return [(track_id, determine_split(i)) for i, track_id in enumerate(track_ids)]
 
 
 def main(known_args: argparse.Namespace, pipeline_args: List[str]) -> None:

basic_pitch/data/datasets/ikala.py

@@ -138,21 +138,17 @@ def process(self, element: List[str], *args: Tuple[Any, Any], **kwargs: Dict[str
 def create_input_data(train_percent: float, seed: Optional[int] = None) -> List[Tuple[str, str]]:
     assert train_percent < 1.0, "Don't over allocate the data!"
 
-    # Test percent is 1 - train - validation
-    validation_bound = train_percent
-
     if seed:
         random.seed(seed)
 
-    def determine_split() -> str:
-        partition = random.uniform(0, 1)
-        if partition < validation_bound:
-            return "train"
-        return "validation"
-
     ikala = mirdata.initialize("ikala")
+    track_ids = ikala.track_ids
+    random.shuffle(track_ids)
+
+    def determine_split(index: int) -> str:
+        return "train" if index < len(track_ids) * train_percent else "validation"
 
-    return [(track_id, determine_split()) for track_id in ikala.track_ids]
+    return [(track_id, determine_split(i)) for i, track_id in enumerate(track_ids)]
 
 
 def main(known_args: argparse.Namespace, pipeline_args: List[str]) -> None:

basic_pitch/data/datasets/medleydb_pitch.py

@@ -0,0 +1,187 @@
+#!/usr/bin/env python
+# encoding: utf-8
+#
+# Copyright 2024 Spotify AB
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import argparse
+import logging
+import os
+import random
+import time
+from typing import Any, Dict, List, Optional, Tuple
+
+import apache_beam as beam
+import mirdata
+
+from basic_pitch.data import commandline, pipeline
+
+
+class MedleyDbPitchInvalidTracks(beam.DoFn):
+    def process(self, element: Tuple[str, str], *args: Tuple[Any, Any], **kwargs: Dict[str, Any]) -> Any:
+        track_id, split = element
+        yield beam.pvalue.TaggedOutput(split, track_id)
+
+
+class MedleyDbPitchToTfExample(beam.DoFn):
+    DOWNLOAD_ATTRIBUTES = ["audio_path", "notes_pyin_path", "pitch_path"]
+
+    def __init__(self, source: str, download: bool) -> None:
+        self.source = source
+        self.download = download
+
+    def setup(self) -> None:
+        import apache_beam as beam
+        import mirdata
+
+        self.medleydb_pitch_remote = mirdata.initialize("medleydb_pitch", data_home=self.source)
+        self.filesystem = beam.io.filesystems.FileSystems()  # TODO: replace with fsspec
+        if self.download:
+            self.medleydb_pitch_remote.download()
+
+    def process(self, element: List[str], *args: Tuple[Any, Any], **kwargs: Dict[str, Any]) -> List[Any]:
+        import tempfile
+
+        import numpy as np
+        import sox
+
+        from basic_pitch.constants import (
+            AUDIO_N_CHANNELS,
+            AUDIO_SAMPLE_RATE,
+            FREQ_BINS_CONTOURS,
+            FREQ_BINS_NOTES,
+            ANNOTATION_HOP,
+            N_FREQ_BINS_NOTES,
+            N_FREQ_BINS_CONTOURS,
+        )
+        from basic_pitch.dataset import tf_example_serialization
+
+        logging.info(f"Processing {element}")
+        batch = []
+
+        for track_id in element:
+            track_remote = self.medleydb_pitch_remote.track(track_id)
+
+            with tempfile.TemporaryDirectory() as local_tmp_dir:
+                medleydb_pitch_local = mirdata.initialize("medleydb_pitch", local_tmp_dir)
+                track_local = medleydb_pitch_local.track(track_id)
+
+                for attr in self.DOWNLOAD_ATTRIBUTES:
+                    source = getattr(track_remote, attr)
+                    dest = getattr(track_local, attr)
+                    os.makedirs(os.path.dirname(dest), exist_ok=True)
+                    with self.filesystem.open(source) as s, open(dest, "wb") as d:
+                        d.write(s.read())
+
+                # will be in temp dir and get cleaned up
+                local_wav_path = "{}_tmp.wav".format(track_local.audio_path)
+                tfm = sox.Transformer()
+                tfm.rate(AUDIO_SAMPLE_RATE)
+                tfm.channels(AUDIO_N_CHANNELS)
+                tfm.build(track_local.audio_path, local_wav_path)
+
+                duration = sox.file_info.duration(local_wav_path)
+                time_scale = np.arange(0, duration + ANNOTATION_HOP, ANNOTATION_HOP)
+                n_time_frames = len(time_scale)
+
+                if track_local.notes_pyin is not None:
+                    note_indices, note_values = track_local.notes_pyin.to_sparse_index(
+                        time_scale, "s", FREQ_BINS_NOTES, "hz"
+                    )
+                    onset_indices, onset_values = track_local.notes_pyin.to_sparse_index(
+                        time_scale, "s", FREQ_BINS_NOTES, "hz", onsets_only=True
+                    )
+                    note_shape = (n_time_frames, N_FREQ_BINS_NOTES)
+                # if there are no notes, return empty note indices
+                else:
+                    note_shape = (0, 0)
+                    note_indices = []
+                    onset_indices = []
+                    note_values = []
+                    onset_values = []
+
+                contour_indices, contour_values = track_local.pitch.to_sparse_index(
+                    time_scale, "s", FREQ_BINS_CONTOURS, "hz"
+                )
+
+                batch.append(
+                    tf_example_serialization.to_transcription_tfexample(
+                        track_id,
+                        "medleydb_pitch",
+                        local_wav_path,
+                        note_indices,
+                        note_values,
+                        onset_indices,
+                        onset_values,
+                        contour_indices,
+                        contour_values,
+                        note_shape,
+                        (n_time_frames, N_FREQ_BINS_CONTOURS),
+                    )
+                )
+        return [batch]
+
+
+def create_input_data(train_percent: float, seed: Optional[int] = None) -> List[Tuple[str, str]]:
+    assert train_percent < 1.0, "Don't over allocate the data!"
+
+    if seed:
+        random.seed(seed)
+
+    medleydb_pitch = mirdata.initialize("medleydb_pitch")
+    track_ids = medleydb_pitch.track_ids
+    random.shuffle(track_ids)
+
+    def determine_split(index: int) -> str:
+        return "train" if index < len(track_ids) * train_percent else "validation"
+
+    return [(track_id, determine_split(i)) for i, track_id in enumerate(track_ids)]
+
+
+def main(known_args: argparse.Namespace, pipeline_args: List[str]) -> None:
+    time_created = int(time.time())
+    destination = commandline.resolve_destination(known_args, time_created)
+    input_data = create_input_data(known_args.train_percent, known_args.split_seed)
+
+    pipeline_options = {
+        "runner": known_args.runner,
+        "job_name": f"medleydb-pitch-tfrecords-{time_created}",
+        "machine_type": "e2-standard-4",
+        "num_workers": 25,
+        "disk_size_gb": 128,
+        "experiments": ["use_runner_v2"],
+        "save_main_session": True,
+        "sdk_container_image": known_args.sdk_container_image,
+        "job_endpoint": known_args.job_endpoint,
+        "environment_type": "DOCKER",
+        "environment_config": known_args.sdk_container_image,
+    }
+    pipeline.run(
+        pipeline_options,
+        pipeline_args,
+        input_data,
+        MedleyDbPitchToTfExample(known_args.source, download=True),
+        MedleyDbPitchInvalidTracks(),
+        destination,
+        known_args.batch_size,
+    )
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    commandline.add_default(parser, os.path.basename(os.path.splitext(__file__)[0]))
+    commandline.add_split(parser)
+    known_args, pipeline_args = parser.parse_known_args()
+
+    main(known_args, pipeline_args)

basic_pitch/data/download.py

@@ -20,11 +20,12 @@
 from basic_pitch.data import commandline
 from basic_pitch.data.datasets.guitarset import main as guitarset_main
 from basic_pitch.data.datasets.ikala import main as ikala_main
+from basic_pitch.data.datasets.medleydb_pitch import main as medleydb_pitch_main
 
 logger = logging.getLogger()
 logger.setLevel(logging.INFO)
 
-DATASET_DICT = {"guitarset": guitarset_main, "ikala": ikala_main}
+DATASET_DICT = {"guitarset": guitarset_main, "ikala": ikala_main, "medleydb_pitch": medleydb_pitch_main}
 
 
 def main() -> None:

tests/data/test_medleydb_pitch.py

@@ -0,0 +1,68 @@
+#!/usr/bin/env python
+# encoding: utf-8
+#
+# Copyright 2024 Spotify AB
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import apache_beam as beam
+import itertools
+import os
+
+from apache_beam.testing.test_pipeline import TestPipeline
+
+from basic_pitch.data.datasets.medleydb_pitch import (
+    MedleyDbPitchInvalidTracks,
+    create_input_data,
+)
+
+
+# TODO: Create test_medleydb_pitch_to_tf_example
+
+
+def test_medleydb_pitch_invalid_tracks(tmpdir: str) -> None:
+    split_labels = ["train", "validation"]
+    input_data = [(str(i), split) for i, split in enumerate(split_labels)]
+    with TestPipeline() as p:
+        splits = (
+            p
+            | "Create PCollection" >> beam.Create(input_data)
+            | "Tag it" >> beam.ParDo(MedleyDbPitchInvalidTracks()).with_outputs(*split_labels)
+        )
+
+        for split in split_labels:
+            (
+                getattr(splits, split)
+                | f"Write {split} to text"
+                >> beam.io.WriteToText(os.path.join(tmpdir, f"output_{split}.txt"), shard_name_template="")
+            )
+
+    for i, split in enumerate(split_labels):
+        with open(os.path.join(tmpdir, f"output_{split}.txt"), "r") as fp:
+            assert fp.read().strip() == str(i)
+
+
+def test_medleydb_create_input_data() -> None:
+    data = create_input_data(train_percent=0.5)
+    data.sort(key=lambda el: el[1])  # sort by split
+    tolerance = 0.01
+    for _, group in itertools.groupby(data, lambda el: el[1]):
+        assert (0.5 - tolerance) * len(data) <= len(list(group)) <= (0.5 + tolerance) * len(data)
+
+
+def test_create_input_data_overallocate() -> None:
+    try:
+        create_input_data(train_percent=1.1)
+    except AssertionError:
+        assert True
+    else:
+        assert False